Creative Workflows on Magicdoor

Magicdoor helps you move from idea to finished creative in one place. Use fast chat models for ideation, switch to the best image models for visuals, and keep everything in one history with shared memory.

Who this is for

• Writers, designers, marketers, founders
• Anyone turning rough ideas into drafts, visuals, and final assets

Why use Magicdoor for creative work

• One interface with multiple top models (chat and images)
• Automatic web search via Perplexity when you need facts
• Image prompt enhancement for stronger visual outputs
• Usage-based costs so exploration stays affordable

Quick setup

• Start a new chat in /chat
• Set tone and preferences in Settings → Base instructions and unit system
• Optionally create an Assistant for your brand voice: see Start with Assistants

Core workflow: idea → drafts → visuals → polish

1. Brainstorm themes and angles
   • Use Claude Sonnet 5 or GPT-5.5 for fast, high-quality ideation
   • Ask for 5–10 directions with target audience and constraints
2. Outline and script
   • Pick one direction and request a short outline
   • Keep paragraphs short and actionable
3. Generate visuals
   • Describe the desired scene, style, and aspect ratio
   • Let Magicdoor enhance your prompt automatically
   • Try multiple image models to compare styles
4. Iterate quickly
   • Ask for 3–5 variations or style tweaks (lighting, palette, framing)
   • Save strong versions; adjust weak points
5. Polish and finalize
   • Tighten copy; ask for punchier headlines or clearer CTAs
   • Export images and copy together

Model picks that work well

• Chat/ideation: Claude Sonnet 5, GPT-5.5, Gemini 3.1 Pro
• Visuals:
  • Seedream 5 Pro: multi-reference generation and editing
  • GPT Image 2.5 Flare: strong prompt interpretation and editing support
  • Flux 2 Pro: another high-quality generation option
  • Recraft V4.1: design-oriented generation

Prompt patterns to try

• For moodboards: “Create 6 concise visual directions for [theme], each with palette, materials, lighting, and reference keywords.”
• For product shots: “Single object on neutral backdrop, soft studio lighting, 3 angles, 4:5, minimal props, realistic texture.”
• For brand styles: “Keep typography modern, high contrast, ample whitespace, color accents in [brand color].”

Cost-aware iteration

• Start with smaller images and fewer variations; scale up once you like a direction
• Keep text prompts short and specific; avoid unnecessary retries
• Usage-based pricing means you pay only for what you generate

Troubleshooting

• Results look generic: add 3–5 concrete visual references or constraints
• Faces or hands look odd: try a different model, change angle, add lighting details
• Colors drift from brand: include hex values and a short palette description

FAQs

Which image model should I start with?

Begin with Google Nano Banana 2 for low-cost iteration. If you need a different look, compare GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, Flux 2 Pro, and Recraft V4.1.

Can I fact-check marketing claims?

Yes. Ask normally—Magicdoor will use Perplexity to search when needed and cite web sources.

How do I keep a consistent brand voice?

Create an Assistant with your brand guidelines and examples, then write inside that Assistant so tone stays consistent.